Remote, US-limited and time-zone-limited jobs in HN "Who is hiring?"

1. Share of all postings (per quarter)

US-limited and time-zone-limited here are shares of all postings, remote or not.

2. Restrictions among remote postings (per quarter)

Denominator: postings classified as remote. Early quarters have only ~80–200 remote postings, so they are noisy.

3. Remote postings by restriction, including overlap (per year)

"No US/TZ limit" means neither was detected. Many of those are still limited to other countries or regions (e.g. "Remote (EU)", "UK only"), which this analysis does not track.

4. Number of postings per year

5. Yearly numbers

Method

Data: every top-level comment (one comment = one posting) in the 187 monthly "Ask HN: Who is hiring?" threads, April 2011 to October 2026, fetched via the Algolia HN API. Deleted comments are not included. 2011 starts in April, 2026 ends in October.
Unit is the posting, not the role. A posting that lists several roles counts once, and is flagged if any part of it matches.
Remote: the role can be done remotely (fully remote or "onsite or remote"). Hybrid roles, "work from home on Fridays" perks and temporary COVID remote ("remote for now") do not count.
US-limited: applicants must be in or from the US (US + Canada / North America also counts): "Remote (US)", "US only", "must be located in the US", "US citizens", "authorized to work in the US". A US office location alone does not count. Lists with other regions ("Remote (US or EU)") do not count.
Time-zone-limited: applicants must be in, or overlap with, specific time zones, including stated preferences: "US time zones", "CET ±3h", "UTC-1 to UTC+3", "overlap with PST", "EST working hours". "Remote (US time zones)" counts as a time-zone limit, not a US limit.
Classification is regex-based (analysis/classify.py).

Validation

The parser was checked by reading postings by hand and comparing them with its labels:

Sample	Postings	Fully correct
Round 1: 8 months 2012–2025, random + keyword postings (used to tune the parser)	112	100 (89%) before fixes
Round 2: 8 other months 2013–2026, random + keyword postings (held out)	112	108 (96%)
Round 3: postings mentioning time zones (used to tune the parser)	40	35 (88%) before fixes
Round 4: postings mentioning time zones (held out)	30	27 (90%)

Remaining errors are mostly missed restrictions with unusual phrasing, not false positives. No time-zone false positives were found in rounds 3–4. Time-zone limits are therefore probably undercounted by roughly 10%. A few postings are genuinely ambiguous (e.g. "onsite, discussion about remote").